5V regulator outputting 5.9V and more.

[floobydust]

MLCC has a huge voltage coefficient of capacitance which makes them useless at the 100uF 5V level. You're lucky to get 22-30uF at room temp in 1206 i.e. TDK C3216JB1A107 datasheet

The ST datasheet graph is even worse
I hate parts like this because a second-source actually ends up being a terrible mistake in production, when supply chain assumes it's the same part number.

[james_s]

You really have to be careful with MLCC, many people are blissfully unaware of how dramatically the capacitance drops with DC across them. I was also not aware of this trait until it bit me a few times, fortunately in one-off hobby projects. It usually works fine to sprinkle on a bunch of 100n decoupling caps, but if you actually try to calculate the needed capacitance it will really get you if you don't account for that.

[Simon]

well my choice of part was dictated by the load dump protection. So a normal electrolytic might be better? will have to check the specs on those caps I have at work. I just need to get this working, do a demo for my course work and move on.

[Simon]

You really have to be careful with MLCC, many people are blissfully unaware of how dramatically the capacitance drops with DC across them. I was also not aware of this trait until it bit me a few times, fortunately in one-off hobby projects. It usually works fine to sprinkle on a bunch of 100n decoupling caps, but if you actually try to calculate the needed capacitance it will really get you if you don't account for that.

Yea, that's why you use 25-50V parts even at 5V

[wraper]

The ST datasheet graph is even worse
I hate parts like this because a second-source actually ends up being a terrible mistake in production, when supply chain assumes it's the same part number.

LOL, it's almost like ESR requirements for ON semi and ST are mutually exclusive.

[Siwastaja]

Yeah, while old-school non-low-drop linear regulators such as 7805 are still fairly OK, avoid most old low-drop regulators, as they tend to have strict requirements for the exact types and values of output capacitors.

Market is full of regulators specified to be stable with ceramic output capacitors; use them whenever possible. Also try to find one which is happy with lower capacitance, allowing smaller solution area and possibly lower total cost. Some are OK with 1uF while others require 10uF.

Using specific types and values of output capacitors in your BOM tends to eat away the price benefit you get from a cheap, old-tech regulator.

[Simon]

I thought it was a newish regulator as it has the load dump capability, in the past I only ever found 5V regulators that needed wrapping in protection circuitry and rarely accepted auto range voltages. Seems like they took and old design and beefed it up but it's still the old design. I will remove it from my parts library, while the input will withstand the auto environment clearly as an overall part it will not survive the full range of conditions.

[MagicSmoker]

I thought it was a newish regulator as it has the load dump capability, in the past I only ever found 5V regulators that needed wrapping in protection circuitry and rarely accepted auto range voltages. 

I would be very wary of trusting any regulator whose datasheet says it can withstand "load dump" unless the manufacturer certifies compliance with ISO 16750-2 (or the older 7637-2) and specifies the minimum series resistance needed to do so.

Good practice with any kind of automotive product is to add at least a 24V nominal TVS diode, preferably in a SMC or larger package, to provide the bulk of load dump protection. That said, whaddya need an LDO for to provide 5V with a 12V nominal input? If it's to maintain operation even during engine cranking that is not something I would entirely entrust to minimizing the dropout voltage; a blocking diode and some bulk capacitance before the regulator would be much more reliable (and give better load-dump tolerance to boot).

[Simon]

TVS is for spikes, I don't know what they mean by load dump buh I take it to mean the surges that should not exceed 40V. I have a 33V TVS

[MagicSmoker]

TVS is for spikes, I don't know what they mean by load dump buh I take it to mean the surges that should not exceed 40V. I have a 33V TVS

"Load dump" is the term for the voltage overshoot from an alternator that occurs when a heavy load is suddenly disconnected (usually the battery while it is being charged, but it could be any heavy load like electric power steering or the like). An "okayish" article on load dump is here:

https://www.electronicdesign.com/power/eliminate-those-automotive-load-dump-circuit-protection-headaches

You definitely want to get a copy of ISO 16750-2 if you are making commercial products for on-vehicle use.

[james_s]

You really have to be careful with MLCC, many people are blissfully unaware of how dramatically the capacitance drops with DC across them. I was also not aware of this trait until it bit me a few times, fortunately in one-off hobby projects. It usually works fine to sprinkle on a bunch of 100n decoupling caps, but if you actually try to calculate the needed capacitance it will really get you if you don't account for that.

Yea, that's why you use 25-50V parts even at 5V

That won't help you. The reduction in capacitance is more closely related to physical package size, higher voltage rating does not really affect this. Pull up some datasheets and examine them closely and you'll see what I mean. This is generally a good habit to get into.

[Simon]

Well what else do you do? i have seen some capacitors with a graph for capacitance versus voltage but few have them.

[james_s]

If you want to use MLCC for DC filtering you have to find a datasheet with that graph and choose the value accordingly. If you can't find the data for the exact capacitor you are considering you ought to be able to find one for a part with the same dielectric. It's a pitfall of this type of capacitor and something to be aware of. Tantalum, electrolytic and dry polymer are other viable options.

[mikerj]

Well what else do you do? i have seen some capacitors with a graph for capacitance versus voltage but few have them.

The decent manufacturers will have these characteristics available somewhere, though finding them can sometimes be difficult IME.  I really like Murata's "SimSurfing" web app that provides numerous characteristic graphs for every part, and also Spice models for them.

[Simon]

this is a kemet part. I thought they were decent.

[james_s]

You could also test a sample if you have an oscilloscope and a resistor. With the addition of a function generator you can lash up a little low pass filter and observe the characteristics under realistic conditions. I think Kemet is decent stuff but I can't speak for their documentation offhand. It's possible that this data is buried in some other document instead of the individual datasheets. As I mentioned previously it's not nearly as widely known or as well publicized as it ought to be. I think a LOT of people get snared in that trap, but in many cases the reduced capacitance doesn't cause obvious problems so it is never noticed.

[floobydust]

MLCC selection is much more dangerous now due to manufacturers "race to the bottom" for smallest size. The push is to go down a size group with the "same" ratings. 1210 obsolete in favour of 1206, 0805 for 0603 etc.
So the voltage coefficient of capacitance effect is much greater now. 0603 is only about 33%of rated C at 5V for a 25V part, whereas 1206 is around 90% - you want the physically biggest part.
TDK MLCC allows you to sort by the capacitance with DC bias, which helps a lot.
Taiyo Yuden also allows DC bias sort and check.

[wraper]

0603 is only about 33%of rated C at 5V for a 25V part, whereas 1206 is around 90% - you want the physically biggest part.

Depends on capacitance for size. Say if you take 4.7uf and 22uF part with same size and 6.3 V rating, there are chances that at 5V actual capacitance is almost the same for both.

[floobydust]

Yes, there seems to be a density where it makes more sense to use a lower value part. Temperature and aging are also worth looking at.
Example 0603 6.3V at 5V:
47uF  gives 9.4uF -80%
22uF  gives 4.4uF -70%
10uF  gives 3.0uF -70%
4.7uF gives 3.3uF -30%

[viperidae]

That 4.7 is an x7r dielectric and the others are x5r. I think that's why it's better

[Simon]

Well 25V on 5V is surely not bad. it's 1206. Yes as soon as i chose X7R all the 0805 parts disappeared from the list.